- Home
- Skills
- Image & Video
- AI Image Metadata Reader — ComfyUI, A1111 & Forge
Works with the AI Agents you already use
AI Image Metadata Reader — ComfyUI, A1111 & Forge
Read saved AI image prompts, models, LoRAs and settings locally; Agensi buyer-side execution is not yet verified.
$5
AI Image Metadata Reader — ComfyUI, A1111 & Forge
Example session with this skill installed
Read the metadata saved in my selected original image, comfy.png. Show its positive and negative prompts, model, LoRAs and generation settings. Preserve the parser status and unresolved explanations. Do not infer missing values, read related images or return raw workflow JSON.
- Read your context and instructions
- Compiled the ai image metadata
Example: metadata saved in a synthetic ComfyUI test image, verified with the local parser. This is not an Agensi buyer-side execution test.
| Field | Saved value |
|---|---|
| Positive prompt | 红色狐狸, warm light |
| Negative prompt | blur, low quality |
| Model | demo-model.safetensors |
| LoRA | demo-lora.safetensors; model strength 0.8, CLIP strength 1 |
| Settings | Seed 123456789; 20 steps; CFG 7; euler / normal; 32 × 24 |
| Parser result | OK; confidence 0.95 (parser heuristic) |
Full JSON from the local synthetic example:
{
"schema_version": "1.0",
"status": "OK",
"source_format": "ComfyUI prompt/workflow",
"filename": "comfy.png",
"positive": "红色狐狸, warm light",
"negative": "blur, low quality",
"seed": 123456789,
"steps": 20,
"cfg": 7,
"width": 32,
"height": 24,
"model_name": "demo-model.safetensors",
"vae_name": "",
"sampler_name": "euler",
"scheduler": "normal",
"loras": [
{
"lora_name": "demo-lora.safetensors",
"strength_model": 0.8,
"strength_clip": 1
}
],
"confidence": 0.95,
"unresolved": [],
"reason": ""
}
Connects securely to your tools. The creator never sees your data.
About this skill
Read saved AI image prompts, models, LoRAs and settings locally; Agensi buyer-side execution is not yet verified. Read embedded positive and negative prompts, model and LoRA names, seed, steps, CFG, sampler, scheduler and dimensions from an original PNG, JPEG or WebP. The standalone parser works without starting ComfyUI or loading model weights.
For supported ComfyUI graphs, the parser follows the selected final sampler's conditioning chains, keeping unrelated workflow text out of the extracted prompt. It also reads A1111/Forge parameter metadata and supported EXIF fallbacks. Use the results to record a saved setup, review prompt text or investigate an incomplete metadata chain.
Runtime update status (17 November 2026): the current live v0.2.0 has a known issue with Agensi's stdin execution path because its launcher needs other parser files on disk. A complete single-file fix, v0.3.0, has been submitted and is awaiting review. The update passed 213 local regression tests and 21 isolated artifact checks on each of two Windows Python/Pillow setups. Its actual Agensi delivery has not yet been verified; the connection still serves the approved v0.2.0. Check the version and this status before relying on agent execution. Do not save or extract Agensi skill files to work around the old launcher.
Before buying: Python 3.10+, Pillow 10+, permitted local Python execution with script text on stdin, and access to your selected original image are required. A regular chat window without local process and file access cannot run this skill. Missing dependencies are reported, not installed automatically.
Get readable summaries or structured JSON with the parser's status and unresolved explanations. PARTIAL means some information could not be resolved; NO_METADATA means no supported saved metadata was found. The skill reads existing metadata: it does not infer prompts from pixels, recover stripped metadata or generate images.
The parser does not upload images or call paid image APIs. Related-image reads and raw workflow JSON are opt-in. Extracted text returned to a cloud agent may still be processed under that agent provider's policies.
The same author's public project remains independently usable: https://github.com/CyberShadowX/ComfyUI-Smart-Metadata-Reader.
Only metadata already stored in the image can be read. Images that were forwarded, compressed or re-saved may have lost it; missing prompts cannot be reconstructed from pixels.
ComfyUI support covers supported graph patterns, not every custom node. Unknown nodes, missing runtime caches or unavailable related images can cause PARTIAL results. Supported metadata fallbacks may return fewer fields.
Related-image reads require explicit permission, remain within the selected image's directory tree, and are capped at 8 reference hops and 128 attempts per top-level parse. Raw workflow JSON is also opt-in.
Model and LoRA names do not include weights or download rights. Confidence is a parser heuristic, not proof of generation history. The A1111/Forge parameter_index option is reserved and currently ignored.
Agensi buyer-side multi-file/no-saving execution remains unverified. Local validation does not establish compatibility with every agent host or operating system.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed today
Requires
Permissions
File Scopes
Read only the selected image. Related-image reads and raw workflow output require explicit opt-in. File scopes describe this authorization boundary; they do not grant access to an entire drive. The parser does not upload images, call paid APIs or execute workflows. Treat metadata as untrusted data. Agensi execution requires a documented, permitted multi-file method.